Actuarial Outpost > CAS CAS Exam 7 Spring 2020 Progress Thread - Now in October!
 Register Blogs Wiki FAQ Calendar Search Today's Posts Mark Forums Read
 FlashChat Actuarial Discussion Preliminary Exams CAS/SOA Exams Cyberchat Around the World Suggestions

 Fill in a brief DW Simpson Registration Form to be contacted when new jobs meet your criteria.

#641
03-05-2020, 04:08 PM
 AbedNadir Member CAS SOA Join Date: Mar 2014 Studying for FCAS Posts: 2,848

Quote:
I enjoy tendies on occasion
__________________
#642
03-05-2020, 04:12 PM
 kadsura Member CAS Join Date: Oct 2016 Studying for the heck of it Posts: 1,017

Quote:
 Originally Posted by AbedNadir I enjoy tendies on occasion
I don't do options at all, but I have A LOT riding on bonds. They've been doing well lately. Too well, in fact I think I need to trim some positions.
__________________
1 2 3 4 S 5 6 7 8 9 V V O O C
#643
03-05-2020, 04:49 PM
 amp019372 Member CAS Join Date: Aug 2017 Location: Columbus, OH Studying for 7 College: B.S. University of TN, M.S. Ohio State Favorite beer: Spotted Cow Posts: 813

Quote:
 Originally Posted by AbedNadir for a single row of data, isn't that beta going to be constant? my intended solution was that since the scale parameter varies by age due to the adjustment, it is larger for that development period than the previous and produced a lower credibility estimate. I'm assuming that the scale parameter in the credibility factor of Verrall could vary instead of remaining constant in the formula I find thinking of these connections helps me remember the material as I read through the syllabus, I know that it's extremely unlikely for them to ask any specific question I come up with, I'm just trying to enhance my understanding
Idk, I don't remember Verrall super well. I don't remember when phi and beta can vary (if it's by row or column). I'll review that more when I loop back around.
__________________
ACAS | 7 | 8 | 9

#644
03-05-2020, 04:50 PM
 amp019372 Member CAS Join Date: Aug 2017 Location: Columbus, OH Studying for 7 College: B.S. University of TN, M.S. Ohio State Favorite beer: Spotted Cow Posts: 813

Quote:
 Originally Posted by kadsura I don't do options at all, but I have A LOT riding on bonds. They've been doing well lately. Too well, in fact I think I need to trim some positions.
I have a 2% interest savings account and I'm just pouring money into it for the time being. After the market drops quite a bit more I'll probably move it.
__________________
ACAS | 7 | 8 | 9

#645
03-05-2020, 04:51 PM
 amp019372 Member CAS Join Date: Aug 2017 Location: Columbus, OH Studying for 7 College: B.S. University of TN, M.S. Ohio State Favorite beer: Spotted Cow Posts: 813

Quote:
 Originally Posted by AbedNadir I enjoy tendies on occasion
Put your whole 401k on calls for KFC, smart play
__________________
ACAS | 7 | 8 | 9

#646
03-05-2020, 09:47 PM
 AbedNadir Member CAS SOA Join Date: Mar 2014 Studying for FCAS Posts: 2,848

Quote:
 Originally Posted by AbedNadir An analyst is implementing a Bayesian model for the BF method as described by Verrall. He is using column factors derived from an ODP bootstrap and he adjusted residuals by development age to have equal variance using the method suggested by Taylor. He is confused that upon inspection of the results, age 4 incremental losses have more weight given to the BF method than the CL method when compared to the previous age. He expected the incremental losses to have increased weight given to the chain ladder method as age increased. Describe what could be causing this result. assume he's just looking at a single row of data with a given BF ultimate estimate
End of section 6 Verrall addresses having a constant phi parameter and it's just done for convenience, so my question seems valid
__________________
#647
03-05-2020, 09:58 PM
 AbedNadir Member CAS SOA Join Date: Mar 2014 Studying for FCAS Posts: 2,848

the ODP model defined by Verrall in the stochastic BF appears to be the same definition as the ODP cross classified from Taylor. The difference appears to be that Verrall's uses MCMC to solve for parameters, which means theorem 3.2 in Taylor may not apply, I think? I am wondering if Verrall's MCMC model with all improper priors would be equivalent to the MLE estimates, in which case they would equal chain ladder by Theorem 3.2?
__________________
#648
03-06-2020, 11:56 AM
 AbedNadir Member CAS SOA Join Date: Mar 2014 Studying for FCAS Posts: 2,848

there are some conflicting messages regarding variability of the reserve. On one hand, Clark says lower variability is better and advocates that cape cod is better for its lower parameter variance. On the other hand, Meyer's demonstrates that models with lower variance did not validate on the Cas loss database and pursues efforts to increase the variability of his estimates, the exact opposite of what Clark was trying to do.

am I missing something or are they really this contradictory?
__________________
#649
03-06-2020, 01:42 PM
 amp019372 Member CAS Join Date: Aug 2017 Location: Columbus, OH Studying for 7 College: B.S. University of TN, M.S. Ohio State Favorite beer: Spotted Cow Posts: 813

Quote:
 Originally Posted by AbedNadir there are some conflicting messages regarding variability of the reserve. On one hand, Clark says lower variability is better and advocates that cape cod is better for its lower parameter variance. On the other hand, Meyer's demonstrates that models with lower variance did not validate on the Cas loss database and pursues efforts to increase the variability of his estimates, the exact opposite of what Clark was trying to do. am I missing something or are they really this contradictory?
My interpretation is that Clark was saying we want to lower parameter variance, not process variance. The parameter variance being lower improves the fit of your model. However the process variance is inherent to the reserving and insurance process, so you wouldn't want to minimize it. The Meyers paper is saying the overall variability of the models were too low, which from my point of view means the process variance was estimated to be lower than it actually is. So with that view, there's not actually a contradiction.

I'm still not at the point where I'm really confident in the material yet though, so I could be wrong.
__________________
ACAS | 7 | 8 | 9

#650
03-06-2020, 01:45 PM
 AbedNadir Member CAS SOA Join Date: Mar 2014 Studying for FCAS Posts: 2,848

Quote:
 Originally Posted by amp019372 My interpretation is that Clark was saying we want to lower parameter variance, not process variance. The parameter variance being lower improves the fit of your model. However the process variance is inherent to the reserving and insurance process, so you wouldn't want to minimize it. The Meyers paper is saying the overall variability of the models were too low, which from my point of view means the process variance was estimated to be lower than it actually is. So with that view, there's not actually a contradiction. I'm still not at the point where I'm really confident in the material yet though, so I could be wrong.
Clark shows that the overall variance is lower for CC due to the trade off being better for decreasing parameter variance and increasing process variance. The discussion by Meyer's is about overall variance as you said. So we have Clark advocating for lower overall variance, and Meyer's trying to get more overall variance
__________________